defect imaging by micromachined ultrasonic air ... - Stanford University

DEFECT IMAGING BY MICROMACHINED ULTRASONIC AIR TRANSDUCERS Sean Hansen, Neville Irani, F. Levent Degertekin, Igal Ladabaum, and B. T. Khuri-Yakub Edward L. Ginzton Laboratory Stanford University Stanford. CA 94305-4085 Abstract-Capacitive

micromachined ultrasonic transduc-

e m (eMUTs) are shown t o have over 100 dB dynamic range in air. This enables fast imaging of internal defects of solid

structures withhigh signal-to-noise ratio. The high dynamic range is t h e result of a resonant structure with a fractional bandwidthlimited toahout 10%. Bettertemporal reslution is required t o differentiate t h e defects in the depth dimension, which demands higher bandwidth devices. In this paper w e present an optimized pulse-echo electronics system for cMUTs in air. Simulations suggest that dynamic ranges in excess of 100 dB are attainable in pulseecho operation using commercially available discrete mmponents. Transmission experiments through aluminum and composite plates verify more than 100 dB dynamic range and demonstrate the ability of cMUTs t o image defects in air at 2.5 MHz. W e also present a variation on cMUT design which improves t h e useful bandwidth of the device, permitting greater depth resolution in pulse-echo imaging.

Fig. 2. Magnified view of cMUT transducer with membrane radii of

50 pm.

and amplifying received echos on the order of tens of microvolts. One suchelectronic system is shown in Fig. 3. In the far left of Fig. 3, the cMUT is modeled as a series RLC circuit in parallelwitha capacitor. These elements are sufficient t o model the fundamental resonance of the INTRODUCTioN transducer, shown in the impedance curve of Fig. 4. ValDue to the large impedance mismatchbetween common ues for the resistors, inductors, and capacitors are found piezoelectric materials and air, conventional piezoelectric by curve-fitting data measured from anetwork analyzer transducers are not very efficient sources of Ultrasound in or data from a simulated impedance curve of the transair [l].While the efficiency can he increased with matchducer for a particular manufacturing process. The resistor ing layers, this improvement often comes at, the expense of of the series RLC is divided into two series resistors. Rrod bandwidth. Recently, several capacitive micromachined ulrepresents the radiationresistance that delivers ultrasound trasonic transducers (cMUTs) have been developed, which models dissipaenergy into air. The second resistor, RI,,,,, are capable of efficient excitation and detection of ultrative losses in the transducer. Simulations and experimental sound in air [2], [3]. As depicted in Fig. 1; a single element data suggest that the value of the loss resistance is approxconsists of a metalized 0.5-1pm thick nitride membrane imately seven times the radiation resistance [4]. suspended above a silicon substrate. Several thousand such Other circuit components include, a coupling capacitor, elements are electrically connected in parallel to make the C,,,,, which isolates the remaining electronics from the transducer, shown in Fig. 2. When the membranes are bitransducerDC bias. A variable inductor, labeled Lt,,, ased with a DC bias voltage, the transducer is capable of in Fig. 3, cancels the transducer's reactive impedance, reefficient excitation and detection of ultrasoundin air. sulting in more efficient Dower deliverv to and from the transducer. More complicated matching networks can be for implemented to transform the transducer impedance maximum power delivery. However, following the transY

_ - - - - -CCDUP I,#".,

I Fig. I. Schematic cross-section

of a single cMUT membrane.

Rrad

L L.ITC9UI".

A

PULSE-ECHO ELECTRONICS SYSTEhl FOR

ChfUT

The use of cMIJTs in monostatic pulse-echo operation requires an electronic system that is capable of handling tone-bursts of tens of volts for t,he transmitted ultrasound

0-7803-4095-7/98/$10.00 0 1998 IEEE

1

M)

1- - 1

r50 V OCBUI far MUT

L

LOWNIIII~

cz:,

m p m r

Fig. 3. A pulse-echo electronics system

1998 IEEE ULTRASONICS SYMPOSIUM - 1003

Fig. 4. Real and imaginary impedance of typical eMUT in air. Fig. 5. Variation in rewived pulse-echo rms voltage signal as a function of R,li.. and R,,,.

mit tonoburst, the transducer radiation resistance, Rlod, changes its role in the circuit from the load impedance to the source impedancewhen it receives an echo. Therefore, in general, one cannot provide a matching network that is capable of matching the transducer for maximum power delivery in both the transmit and receive cases. A single tuning inductor is an easily implemented compromise for the system. Thesystemtransmitsanultrasoundtone-burst when driven by a large '20 V amplitude AC signal source and some fraction of the available power is dissipated in the radiation resistance as ultrasound in air. The resistor, Rdiss, in front of the amplifier of Fig. 3 attenuates theAC signal in front of the protectiondiodes, protecting the amplifier from large voltages a t its input. During the transmit tone burst, it is advantageous touse a large value of Rdi.. to minimize the amount of generator power that it shunts away from the transducer. When the system receives an echo, the AC signal generator is shorted and all of the diodes are effectively off since the received signal amplitude is significantly smaller than the diodes' built-in potentials. The received signal generaof as an AC voltage sourcewith a source tor can be thought resistance of Rrod. Since forms a voltage divider with the input of the amplifier, R,,,,, it is advantageous to use a small value for Rdlaa for the maximum received signal. From the preceding discussion, it appears there is an optimum set of values for Rdias and Rampwhich balances the losses during the transmit andreceive phases of pulseecho operation. For a transducer with the impedance curve shown in Fig. 4 , the dependence on the received voltage signal at the inputof the amplifier for a fixed AC generator voltage is shown in Fig. 5. Typically, the received voltage is more sensitive to the value of Rdiasthan to values of R,,,. Since it is not practicalt o design an amplifier with arbitrary input impedances for use with particular transducers, one may insert a transformer in front of an amplifier that has a known input impedance. Once values for Rdlsaand the transformer ratiofor Ram, are selected, complete signal and noise analysis on t,he full system is possible using SPICE circuitsimulation.This

1004 - 1998 IEEE ULTRASONICS SYMPOSIUM

analysis accounts for non-ideal signal losses through parasitic capacitances and for all thermal noise sources. The transducer's own thermal noise is accurately represented by Rrnd and RI,,, [4]. For the amplifier input, the circuit uses the AD600, a commerciallyavailable low-noise amplifier from Analog Devices with an input noise voltage of 1.4 n V J m . Since the AD600 has an input resistance of 100 Cl, a wide range of effective input resistances is available by selecting the appropriate transformer ratio at input to the amplifier. The simulated noise bandwidth is 1 MHz around the transducer resonance: which is easily achieved by filtering the output of the amplifier. Figure 6 shows ............................................

~

-

~~~~~~~~

~~~~~~~~

!'. l \

.

.~

the squared noise voltage in V'/Hz a t t h e amplifier imput, including a 1.4 n V J m noise source in front of the AD600 amplifier. The square-root of the integral of this noise curve yields a noise voltage at the input of the now noiseless amplifier of 1.057 pVrms. Since the received echo is 0.216 Vrms when all of the radiated power from R,,,L is available for reception, the pulse-echo system has a dy-

namic range of 106 dB. A transmission experiment,which uses separate transmit and receive circuits. is capable of a larger dynamic rangeof operation since it is possible to independently match both the transducer and receiver for maximum available power delivery. Furthermore, there is no need for dissipative elements, such as Rd,.., or diode protectioncircuits since the sensitive amplifier is isolated from the transmit signal. When an analysis with the same noise parameters of the preceding example is applied to a transmission experiment, in which both transmitting and receiving circuits are perfectly matched, the simulated dynamic range is 115 dB. These large dynamic ranges, in excess of 100 dB for both pulse-echo and transmission operation,show the applicability of cMUTs for NDE and defect imaging. Further design improvements which reduce the internal loss of the transducer, modeled by RI,,,,, could increase the dynamic range an additional 20-30 dB.

Fig. S. Amplitude image of composite prior to impact damage.

DEFECTIMAGINGRESULTS The simulations of the previous section suggest that the dynamic range of cMUTs is adequate to transmit through materials such as metal and composite plates in air, without the need for coupling fluids. Figure 7 shows the received tone from bistatic transmission through a four-layer carbon fiber composite plate at normalincidence. By varying the amplification of the received signal with and without the sample, we estimate that the composite plate and its interfaces with air attenuate the ultrasound signal by 68 dB. With an additional 6 d B of attenuation due to the air gapat 2.3 MHz,and a 29 dB signal-to-noise ratio, Fig. 7 demonstrates a dynamic range of 103 dB.

Amplitude image of composite after impactdamage in two places.

Fig. 9.

plate are evident by regions of light and dark. Figure 9 is produced by scanning the same sectionof composite plate after the composite was struck in two places, which appear as dark spots in the figure. The leftmost damage spot is not optically visible from the surface_ hutis clearly present in the ultrasound image. The image in Fig. 10 shows the amplitude variationsin a 3 mm thick aluminum plate that hasa 0.5 mm deep pattern milled on the underside of the plat,e. The dynamic range of the the system in this experimentis also about 103 dB, with 82 dB of loss through the aluminum plate at 2.3 MHz. Fig. 7. Receivedvoltage signal through composite plate

This transmission experiment did not take advantage of any matching networks to boost the power delivery to and from the t,ransducers. Nonetheless, its 29 dB signal t o noise ratio, without signal averaging, is adequate to image damage in the composite plate. Figure 8 shows an image produced by linearly gray-scaling the amplitudeof the received signal over a section of composite. Density variations in the

DYNAMIC FREQUENCY TUNING O F cMUTs Inmany ultrasound applications such as pulse-echo imaging, it is desirable t o have a wide band transducer t,o allow shorter excit,ation tone-bursts. This permits better resolution of reflections in time from varying depths in the material. Although the resonant structure of our current cMUT devices for airresults in a large dynamic range, they are inherently narrow band devices with fractional hand-

1998 IEEE ULTRASONICS SYMPOSIUM - 1005

Fig. 10. Amplitude image of milled pattern in aluminum plate.

Fig. 12. Experimental test of resonance shift in an immersion chlUT.

widths of less than 10%. In principle, a narrow band transducer can provide the same information as a wide hand transducer if it can proWe vide a signal output over the same frequency range. propose a method by which the resonant frequency of the cMUT can he altered quickly and accurately using pier+ electricactuation.Figure 11 illustratesonemethod of applyingstress which hasbeentestedonanimmersion cMUT. Figure 12 shows theshift in a transducer's resonant frequency, as measured on a network analyzer. Computer

l

Rlgld

holder

I

Fig. 11. A possible piezoelectric actuation scheme for dynamic controlofcMUT resonance frequency.

simulations on an immersion transducer, shown in Fig. 13, suggest that a wide variation in frequency shifts may be possible withcMUTs.Signals collected from a dynamically strained cMUT in pulse-echo operation can he analyzed in the frequency domain instead of the conventional t,ime domain analysis for a wide band transducer.

CONCLUSION Electronic circuit simulation suggests that a pulse-echo imaging system using cMUTsis capable of dynamic ranges in excess of 100 dB. Exnerimental data from transmission experimentsdemonstrates 103 dBdynamicrangewithout the use of matching networks or signal averaging. This permits defect imaging in aluminum and composite plates. Further development of c ~ [ is~ necessary ~ s to improve theirability totemporally resolve defects in a pulse-echo configuration. However, simulationsandpreliminary experimental results demonstrate the feasibility of dynamically altering the membrane resonance t o achieve a wide band resoonse.

1006 - 1998 IEEE ULTRASONICS SYMPOSIUM

ACKNOWLEDGMENTS This work is supported by WPAFB and the US Office of Naval Research.

REFERENCES [l] W.A Grandia and C.M. Fortunka, "Nde applications of air^ coupled UltraSoniC transducers,'. Seattle, WA, November 1995, IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society, pp. 697-709. [2] D.W. Schindel and D.A. rhtchins,"Applications of micromachined capac.itance transducer3 in air-coupled ultrasonics and nondestructive evaluation," IEEE %ns. on [fltrosonics, Ferne l e c t k s ond Frequency Contml, vol. 42, no. 1,pp. 51-23, January 1995. 131 M. I. Haller and B. T.Kliuri~Yakub. "A surface micromachined electrostatic UltraSOniC air transducer," IEEE Pons. O l l Ultmsonies, Ferroelectrics and Frequency Contrvl, vol. 4 3 , no. 1, pp. January 1996, 141 I. Ladabaum, X. C. Jin, H. T. Soh, A . Atalar, and B. 'r. KhuriYakub, "Surface micromacbined capacitive Ultrasonic transdoce m ? IEEE ??om. an Ultrasonics, Ferroelectrics and Frequency Control, 45, no, 3 , pp, 678-69~, hfay 19923,

..